Fibre characterization: CD, PMD, spectral attenuation, ORL, latency

A loss test says whether a fibre passes light. It does not say whether the fibre can carry 10G over 80 km, forty DWDM channels or a 400ZR signal — those questions depend on chromatic dispersion, polarization mode dispersion, attenuation across the whole band, reflections and sometimes latency. Measuring them is fibre characterization. This page says when it is needed, which test answers which question, how each measurement is made, what the numbers must be for each service, and what the results reveal about old or unknown fibre before money is spent on transceivers.

When characterization is needed

Situation	Tests beyond the standard loss test and OTDR
New route, ≤ 10G, ≤ 40 km, modern G.652.D	none — Tier 1 loss at 1310/1550 nm and a bidirectional OTDR are enough (Testing & measurement)
10G over 60–80 km, 10G DWDM direct detect	chromatic dispersion (the limit is 1 000–1 600 ps/nm), ORL
Legacy 40G NRZ or DPSK direct detect	PMD (very sensitive), CD with compensation planning
100G+ coherent, 400ZR/ZR+	loss and ORL; CD to identify the fibre type (G.653/G.655 change the non-linear plan); PMD to confirm it is within the DSP window; spectral attenuation for C-band flatness; latency if the service asks
Dark fibre lease or acceptance of a leased pair	the full suite, both directions, three wavelengths, plus fibre type identification, reflectance map and endface images — the fibre's passport
Old fibre (installed before 2000) before an upgrade	PMD first (it can be the killer), CD to identify the type, spectral attenuation for the water peak and hydrogen (Reliability & ageing)
CWDM on all 18 channels	spectral attenuation 1 270–1 610 nm — the 1383 nm water peak decides three channels
PON overlays (1490/1577/1625 nm)	attenuation at those wavelengths, ORL ≥ 32 dB (Reflections & return loss)
Fronthaul, time synchronization	latency and the length asymmetry between the two fibres of a pair — PTP accuracy depends on it

The tests

Test	Method	Result	Typical instrument
Spectral attenuation	broadband source and optical spectrum analyzer, swept tunable laser with a power meter, or a multi-wavelength OTDR (1310/1383/1490/1550/1625/1650 nm)	dB/km versus wavelength over 1 260–1 650 nm; the water peak at 1383 nm, hydrogen, bend loss rising toward 1 625 nm, C-band flatness	OSA with source, or a characterization OTDR
Chromatic dispersion (CD)	phase-shift or differential phase-shift (IEC 60793-1-42), or OTDR-based four-wavelength method fitting the dispersion curve	D(λ) in ps/(nm·km), zero-dispersion wavelength λ₀, slope S₀, total dispersion of the link in ps/nm (Dispersion & bandwidth)	CD analyzer or OTDR module with a remote source
Polarization mode dispersion (PMD)	interferometric (TINTY/GINTY), fixed-analyzer wavelength scanning, Jones-matrix; distributed P-OTDR to find the bad section	total PMD in ps (mean DGD) and the coefficient in ps/√km; a statistical quantity — repeat and note temperature	PMD analyzer with a polarized broadband source at the far end
Optical return loss and reflectance	OCWR or OTDR	ORL in dB, reflectance per event	ORL meter, OTDR
Insertion loss	light source and power meter at every service wavelength, both directions (Tier 1)	dB per fibre	OLTS
Latency and length	OTDR length, or a latency tester with a far-end loopback (round trip ÷ 2)	µs and metres; 4.9 µs/km; asymmetry between the fibres of a pair from the length difference at the same index	OTDR, Ethernet tester with latency function
Fibre type identification	from λ₀ and D(1550) of the CD test, backscatter steps at splices, records	G.652 / G.653 / G.655 / G.654 / G.657, era of manufacture (Fibre types)	CD analyzer plus OTDR
Multimode modal bandwidth	differential mode delay / effective modal bandwidth on a sample	OM grade confirmation	laboratory; rarely in the field

Limits per service

Typical figures; the exact values come from the transceiver or the standard.

Service	Chromatic dispersion tolerance	PMD tolerance (mean DGD)	Loss / ORL
1G, 10G up to 40 km (LR, ER)	not limiting on G.652 (40 km × 17 = 680 ps/nm, limit ≈ 1 000–1 600)	≤ 10 ps (10 % of the 100 ps bit)	budget 6–15 dB; ORL ≥ 12 dB
10G 80 km (ZR) and 10G DWDM direct detect	≤ 1 600 ps/nm without compensation; 80 km × 17 = 1 360 ps/nm — tight	≤ 10 ps	budget 23–24 dB; ORL ≥ 24 dB
Legacy 40G NRZ / DPSK	≤ 60–100 ps/nm — needs dispersion compensation	≤ 2.5 ps	ORL ≥ 24 dB
100G coherent DP-QPSK	± 50 000 ps/nm (thousands of km)	25–30 ps and more	OSNR-limited; ORL ≥ 24 dB (Coherent long haul)
400ZR (OIF)	± 2 400 ps/nm (≈ 120 km of G.652)	≈ 10 ps	unamplified: ≤ 11 dB loss (≈ 40 km); amplified: 120 km with OSNR ≥ 26 dB
400ZR+, 800ZR	± 20 000 … ± 50 000 ps/nm	tolerant	OSNR-limited
CWDM 10G, 18 channels	as 10G per channel	≤ 10 ps	attenuation per channel including the water peak (CWDM mux & budget)
GPON / XGS-PON	not limiting	not limiting	class budget (B+ 28, C+ 32, N1 29, N2 31, E1 33, E2 35 dB); ORL ≥ 32 dB; reflectance ≤ −35 dB

PMD arithmetic: mean DGD = PMD coefficient × √L. Modern G.652.D at ≤ 0.1 ps/√km gives 1 ps over 100 km — irrelevant; 1980s fibre at 1–2 ps/√km gives 10–20 ps over 100 km — the end of 10G NRZ on that route unless coherent is used (Formulas & calculations).

Reading the results: fibre identity and age

Measured	Conclusion
λ₀ = 1 300–1 324 nm, D(1550) = 16–18 ps/(nm·km)	G.652; low 1383 nm peak → G.652.C/D, high → G.652.A/B
D(1550) = 2–6 ps/(nm·km), λ₀ around 1 450–1 520 nm	G.655 non-zero dispersion-shifted; check the sign; four-wave mixing needs attention in dense DWDM (Non-linear effects)
D(1550) ≈ 0	G.653 dispersion-shifted — no C-band DWDM without unequal channel spacing or the L-band
D(1550) = 18–22 ps/(nm·km), attenuation ≈ 0.17 dB/km	G.654 pure-silica-core
PMD coefficient > 0.5 ps/√km	fibre from before about 1995; 10G above 100 km is risky; coherent is fine
Water peak at 1383 nm above 0.5 dB/km	no E-band CWDM (1 371–1 411 nm channels); everything else usable
Attenuation slope rising against the baseline, worst at 1 383 and above 1 550 nm	hydrogen ageing or water — a section to replace
Loss rising steeply toward 1 625–1 650 nm	macro- or microbending along the route
Length asymmetry of a pair > 10–20 m	PTP timing error of 50–100 ns unless compensated

Procedure and reporting

Step	Practice
Setup	instruments at both ends with a remote unit or a second crew; reference cords inspected and cleaned (Endface inspection & cleaning); the fibre dark — CD and PMD cannot be measured through a live fibre
Conditions	note date, time and temperature — PMD changes with them; test all fibres of the pair and the spares
Time	30–60 minutes per fibre for the full suite with an experienced crew
Deliverables	per fibre: loss at all wavelengths both directions, ORL, OTDR traces both directions (.sor), CD curve with λ₀ and total dispersion, PMD, spectral attenuation curve, latency and length, reflectance map, endface images; pass/fail against the intended service
Storage	native instrument files with settings, not only PDFs (Documentation & labelling)
Standards	IEC 61280-4-2 (attenuation and ORL of installed single-mode links), IEC 61280-4-4 (PMD of installed links), IEC 61280-4-1 (installed multimode links), IEC 60793-1-42 (CD), TIA FOTP-175 (CD), ISO/IEC 14763-3 and TIA-568.3-D (acceptance) (Standards map)

Dark fibre acceptance checklist: fibre type stated and confirmed by CD; loss per km at 1310/1550/1625 within the offer; every splice ≤ 0.1 dB and every connector ≤ 0.5 dB bidirectionally; ORL ≥ 27 dB (≥ 32 dB for PON); PMD within the service limit; spectral curve without a hydrogen signature; length and latency recorded; documentation and coordinates handed over.

In CodingBox

A coherent module characterizes the fibre for free once it is lit: CMIS VDM reports chromatic dispersion, DGD, PDL, OSNR and the polarization change rate as the DSP sees them. CodingBox shows which of those observables a module advertises and reads its DDM pages, so the commissioning report of the fibre and the module's own view of the same link can be compared — a disagreement points at a patch, a dirty connector or a fibre swap after the survey (VDM & FEC metrics, Check transceiver).